skill-mv/scripts/pipeline-b.mjs

// scripts/pipeline-b.mjs — 产线B 编排：AI 口播讲解视频端到端
//
// 口播稿 → TTS 配音 → 人物讲解画面 → 场景切换 B-roll → 剪辑合成成片
//
// 与产线A 共用同一套底层模块（chat / image / video / tts / ffmpeg），只是编排不同。

import { existsSync, writeFileSync, mkdirSync } from 'node:fs';
import { join } from 'node:path';
import {
  parseArgs, readJSON, writeJSON, paths, ensureWork, log, humanSize, chatJSON,
} from '../lib/index.mjs';
import { buildScriptPrompt, DEFAULT_ANCHOR } from '../lib/prompts.mjs';
import { buildStoryboardFromScript, renderStoryboardMd } from './storyboard.mjs';
import { generateKeyframes } from './keyframe.mjs';
import { generateScenes } from './scene.mjs';
import { generateClips } from './video.mjs';
import { assemble } from './assemble.mjs';
import { getProvider } from '../lib/providers.mjs';
import { printReport } from './pipeline-a.mjs';

const SYSTEM = `你是资深知识类短视频编导兼口播稿写手，稿件会被 AI 语音合成（TTS）直接朗读。
只输出 JSON，不要任何解释文字与 markdown 围栏。`;

const SCHEMA = `严格输出如下 JSON（字段名不可改）：
{
  "title": "标题（≤20字，有吸引力）",
  "topic": "主题",
  "durationSec": 90,
  "hook": "开场 3 秒钩子原句",
  "cta": "结尾行动号召原句",
  "hostAnchor": "英文，讲者外貌锚点（性别/年龄/发型/服装/气质），逐镜头复述用",
  "sections": [{"name":"段落名","text":"该段口播原文"}],
  "fullText": "完整口播稿（各段拼接，用于 TTS 一次性合成）"
}`;

/** 生成口播稿 */
export async function buildOralScript(topic, opts = {}) {
  const { durationSec = 90, audience = '', tone = '专业但口语化', points = [], model } = opts;
  const user =
    buildScriptPrompt({ topic, durationSec, audience, tone, points }) + '\n\n' + SCHEMA;
  const res = await chatJSON({
    model,
    system: SYSTEM,
    messages: [{ role: 'user', content: user }],
    temperature: 0.8,
    max_tokens: 4096,
  });
  const parsed = res.json;
  if (!parsed.fullText) {
    parsed.fullText = (parsed.sections || []).map((s) => s.text).join('\n');
  }
  parsed.hostAnchor = parsed.hostAnchor || DEFAULT_ANCHOR.talk;
  parsed._meta = { model: res.model, createdBy: 'fmode-mv/pipeline-b', topic };
  return parsed;
}

export async function runPipelineB(topic, opts = {}) {
  const {
    scriptFile = '',
    duration = 90,
    aspect = '16:9',
    shots = 8,
    limit = 0,
    outFile = join(paths.out, 'talk.mp4'),
    voice = '',
    music = '',
    skipTts = false,
    skipVideo = false,
    audience = '',
    points = [],
    // v0.2 新增
    characterSheetFile = '',
    skipCharacter = false,
    characterSteps = 4,
    single = false,
    tier = '1k',
    videoMode = 'auto',
  } = opts;

  ensureWork();
  const steps = [];
  const mark = (name, status, detail) => steps.push({ name, status, detail });

  // ---------- 1. 口播稿 ----------
  log.step('[1/8] 口播稿');
  let script;
  const scriptFileOut = join(paths.work, 'script.json');
  if (scriptFile) {
    const { readFileSync } = await import('node:fs');
    const text = readFileSync(scriptFile, 'utf8').trim();
    script = { title: topic || '口播讲解', topic, fullText: text, sections: [{ name: '全文', text }], hostAnchor: DEFAULT_ANCHOR.talk };
    mark('口播稿', 'provided', `读取 ${scriptFile}（${text.length} 字）`);
    log.ok(`读取 ${scriptFile}`);
  } else if (existsSync(scriptFileOut)) {
    script = readJSON(scriptFileOut);
    mark('口播稿', 'reused', `沿用 work/script.json《${script.title}》`);
    log.ok(`沿用已有《${script.title}》`);
  } else {
    script = await buildOralScript(topic, { durationSec: duration, audience, points });
    mark('口播稿', 'ok', `${script.fullText.length} 字 / 约 ${Math.round(script.fullText.length / 3.6)}s`);
    log.ok(`《${script.title}》（${script.fullText.length} 字，约 ${Math.round(script.fullText.length / 3.6)} 秒）`);
  }
  writeJSON(scriptFileOut, script);
  writeFileSync(join(paths.work, 'script.md'), `# ${script.title}\n\n${script.fullText}\n`, 'utf8');

  // ---------- 2. TTS 配音 ----------
  log.step('[2/8] TTS 配音');
  let voiceFile = voice && existsSync(voice) ? voice : null;
  if (voiceFile) {
    mark('TTS 配音', 'provided', voiceFile);
    log.ok(`使用指定配音：${voiceFile}`);
  } else if (skipTts) {
    mark('TTS 配音', 'skipped', '--skip-tts');
    log.warn('按参数跳过');
  } else {
    const p = getProvider('tts');
    const pr = await p.probe({});
    if (!pr.available) {
      mark('TTS 配音', 'unavailable', pr.detail.split('\n')[0]);
      log.warn(`通道不可用：${pr.detail}`);
      log.dim('替代：pip install edge-tts（免费本地）/ FEME_TTS_ENDPOINT 对接外部服务 / --voice 指定现成音频');
    } else {
      try {
        const { synthesize } = await import('./tts.mjs');
        const out = join(paths.audio, 'voiceover.mp3');
        const r = await synthesize(script.fullText, { outPath: out, voice: opts.voiceName });
        voiceFile = r.outPath;
        mark('TTS 配音', 'ok', `${r.provider} → ${voiceFile}`);
        log.ok(`${r.provider} → ${voiceFile}`);
      } catch (e) {
        mark('TTS 配音', 'failed', e.message.split('\n')[0]);
        log.err(`失败：${e.message.split('\n')[0]}`);
      }
    }
  }

  // ---------- 3. 分镜 ----------
  log.step('[3/8] 讲解分镜');
  const sb = await buildStoryboardFromScript(script.fullText, {
    durationSec: duration,
    aspect,
    shotCount: shots,
    hostAnchor: script.hostAnchor || '',
  });
  sb.title = sb.title || script.title;
  writeJSON(paths.storyboard, sb);
  writeFileSync(join(paths.work, 'storyboard.md'), renderStoryboardMd(sb), 'utf8');
  mark('讲解分镜', 'ok', `${sb.shots.length} 镜 / ${sb.totalDuration}s`);
  log.ok(`${sb.shots.length} 个镜头，总时长 ${sb.totalDuration}s`);

  // ---------- 4. 人物资产（④） ----------
  log.step('[4/8] 人物资产多视图（即梦）');
  const jobs = limit ? sb.shots.slice(0, limit) : sb.shots;
  const sbLimited = { ...sb, shots: jobs };
  let characterSheet = characterSheetFile || '';
  if (characterSheet && existsSync(characterSheet)) {
    mark('人物资产', 'provided', characterSheet);
    log.ok(`使用指定人物多视图：${characterSheet}`);
  } else if (skipCharacter) {
    mark('人物资产', 'skipped', '--skip-character');
    log.warn('按参数跳过（跨镜头可能换脸）');
  } else {
    const jp = getProvider('jimeng');
    const jr = await jp.probe({});
    if (!jr.available) {
      mark('人物资产', 'unavailable', jr.detail.split('\n')[0]);
      log.warn(`即梦通道不可用：${jr.detail}`);
    } else {
      try {
        const { buildCharacterSheet } = await import('./character-sheet.mjs');
        const cs = await buildCharacterSheet({
          description: sb.appearanceAnchor || script.hostAnchor || DEFAULT_ANCHOR.talk,
          outDir: paths.out,
          limit: characterSteps,
          single: !!single,
        });
        const sel = cs.steps.find((s) => s.step === '4-多视图' && s.file) || cs.steps.filter((s) => s.file).pop();
        characterSheet = sel?.file || '';
        mark('人物资产', cs.cost.total > 0 ? 'ok' : 'reused', `${cs.steps.length} 步，¥${cs.cost.total.toFixed(2)}`);
        log.ok(`完成 ${cs.steps.length} 步，花费 ¥${cs.cost.total.toFixed(2)}${characterSheet ? ` → ${characterSheet}` : ''}`);
      } catch (e) {
        mark('人物资产', 'failed', e.message.split('\n')[0]);
        log.err(`失败：${e.message.split('\n')[0]}`);
      }
    }
  }

  // ---------- 5. 场景图 + 关键帧（⑤） ----------
  log.step('[5/8] 场景图（无人）→ 关键帧（有人）');
  const jp2 = getProvider('jimeng');
  const jr2 = await jp2.probe({});
  if (jr2.available) {
    const sc = await generateScenes({
      storyboard: sbLimited,
      characterSheet,
      sceneDir: join(paths.work, 'scene'),
      frameDir: paths.frames,
      tier,
    });
    mark('场景图+关键帧', sc.failed ? 'partial' : 'ok', `成功 ${sc.ok} / 失败 ${sc.failed} / ¥${sc.cost.total.toFixed(2)}`);
    log.ok(`成功 ${sc.ok} 张 → ${paths.frames}（¥${sc.cost.total.toFixed(2)}）`);
  } else {
    log.warn(`即梦通道不可用，退回旧 image 通道：${jr2.detail}`);
    const kf = await generateKeyframes({ storyboard: sbLimited, outDir: paths.frames, aspect });
    mark('场景图+关键帧', kf.failCount ? 'partial' : 'ok', `（旧通道）成功 ${kf.okCount} / 失败 ${kf.failCount}`);
    log.ok(`（旧通道）成功 ${kf.okCount} 张 → ${paths.frames}`);
  }

  // ---------- 6. 场景 B-roll / 讲解画面（⑥） ----------
  log.step('[6/8] 图生视频（即梦）');
  let clipResult = { results: [], mode: 'skipped', cost: { total: 0 } };
  if (skipVideo) {
    mark('视频片段', 'skipped', '--skip-video');
    log.warn('按参数跳过');
  } else {
    clipResult = await generateClips({ storyboard: sbLimited, mode: videoMode });
    const ok = clipResult.results.filter((x) => x.status === 'ok').length;
    const degraded = clipResult.results.filter((x) => x.degraded).length;
    mark('视频片段', clipResult.mode === 'jimeng' ? 'ok' : 'fallback', `${clipResult.mode === 'jimeng' ? '即梦真通道' : 'Ken Burns 兜底'}：${ok} 段${degraded ? `（${degraded} 段降级）` : ''}，¥${(clipResult.cost?.total || 0).toFixed(2)}`);
    log.ok(`${clipResult.mode === 'jimeng' ? '即梦真通道' : '兜底'}：${ok} 段，花费 ¥${(clipResult.cost?.total || 0).toFixed(2)}`);
  }

  // ---------- 6. 唇形（可选） ----------
  const lipShots = sbLimited.shots.filter((s) => s.lipSync);
  if (lipShots.length) {
    log.step('[7/8] 讲者唇形驱动');
    const p = getProvider('lipsync');
    const pr = await p.probe({});
    if (pr.available && voiceFile) {
      try {
        const { driveLip } = await import('./lipsync.mjs');
        for (const s of lipShots) {
          const face = join(paths.frames, `shot-${String(s.id).padStart(2, '0')}.png`);
          if (existsSync(face)) await driveLip(s, { faceImage: face, audio: voiceFile });
        }
        mark('唇形驱动', 'ok', `${lipShots.length} 个镜头`);
      } catch (e) {
        mark('唇形驱动', 'failed', e.message.split('\n')[0]);
      }
    } else {
      mark('唇形驱动', 'unavailable', `${lipShots.length} 个讲者镜头未对口型`);
      log.warn(`通道不可用，${lipShots.length} 个讲者镜头将不对口型（观感影响有限）`);
    }
  }

  // ---------- 7. 剪辑合成 ----------
  log.step('[8/8] 剪辑合成');
  let final = null;
  try {
    const r = await assemble({
      storyboard: sbLimited,
      clipsDir: paths.clips,
      outFile,
      voice: voiceFile || undefined,
      bgm: music || undefined,
      bgmVolume: music ? 0.18 : 0,
      beatSync: false,
    });
    final = r.outFile;
    mark('剪辑合成', 'ok', `${r.info.duration.toFixed(1)}s / ${humanSize(r.info.size)}`);
  } catch (e) {
    mark('剪辑合成', 'failed', e.message.split('\n')[0]);
    log.err(`失败：${e.message.split('\n')[0]}`);
  }

  const report = {
    pipeline: 'B',
    topic,
    title: script.title,
    aspect,
    shots: sb.shots.length,
    final,
    steps,
    generatedAt: new Date().toISOString(),
  };
  writeJSON(join(paths.work, 'pipeline-b.run.json'), report);
  printReport('产线B · AI 口播讲解视频', report);
  return report;
}

export async function main(argv = process.argv.slice(2)) {
  const { _, flags } = parseArgs(argv);
  const topic = _.join(' ').trim();

  if (flags.help || (!topic && !flags.script)) {
    console.log(`
fmode-mv talk — 产线B：AI 口播讲解视频端到端编排

用法:
  fmode-mv talk "讲解主题"
  fmode-mv talk "主题" --points "要点1,要点2" --duration 90
  fmode-mv talk --script my-script.md          # 用现成口播稿

流程:
  口播稿 → TTS配音 → 讲解分镜 → 关键帧 → B-roll/讲解画面 → 唇形(可选) → 剪辑成片

选项:
  --script <file>    现成口播稿（md/txt），跳过自动撰写
  --duration <s>     目标时长（默认 90）
  --points "a,b,c"   必须覆盖的要点（逗号分隔）
  --audience "..."   受众描述，影响语气与用词
  --shots <n>        镜头数（默认 8）
  --limit <n>        只处理前 n 镜（试跑省钱）
  --aspect 16:9|9:16|1:1
  --voice <file>     用现成配音音频（跳过 TTS）
  --voice-name <n>   TTS 音色名
  --music <file>     轻背景乐
  --skip-tts / --skip-video
  --out <file>       成片输出（默认 out/talk.mp4）

⚠️ 通道现状：TTS 平台未开通（本地装 edge-tts 即可用，或用 --voice 指定现成音频）；
   ✅ 画面走即梦（人物资产 / 场景图 / 关键帧 / 图生视频）实测可用；
   ❌ 唇形驱动未开通 → 需要真人对口型时改用 fmode-mv jimeng digital-human。

v0.2 新增选项：
   --character <file>   指定现成的人物多视图（跳过自动生成人物资产）
   --skip-character     跳过人物资产（省钱；跨镜头可能换脸）
   --char-steps <1-4>   人物资产跑到第几步（默认 4=多视图）
   --single             人物资产每步只出 1 张（跳过择优）
   --tier 1k|2k         场景/关键帧尺寸档（默认 1k）
   --video-mode auto|jimeng|still  视频通道（默认 auto）
`);
    return 0;
  }

  await runPipelineB(topic || '口播讲解', {
    scriptFile: typeof flags.script === 'string' ? flags.script : '',
    duration: Number(flags.duration) || 90,
    aspect: flags.aspect || '16:9',
    shots: Number(flags.shots) || 8,
    limit: Number(flags.limit) || 0,
    outFile: String(flags.out || join(paths.out, 'talk.mp4')),
    voice: typeof flags.voice === 'string' ? flags.voice : '',
    music: typeof flags.music === 'string' ? flags.music : '',
    voiceName: typeof flags['voice-name'] === 'string' ? flags['voice-name'] : undefined,
    skipTts: !!flags['skip-tts'],
    skipVideo: !!flags['skip-video'],
    characterSheetFile: typeof flags.character === 'string' ? flags.character : '',
    skipCharacter: !!flags['skip-character'],
    characterSteps: Number(flags['char-steps']) || 4,
    single: !!flags.single,
    tier: flags.tier || '1k',
    videoMode: flags['video-mode'] || 'auto',
    audience: flags.audience || '',
    points: flags.points ? String(flags.points).split(',').map((s) => s.trim()).filter(Boolean) : [],
  });
  return 0;
}

if (import.meta.url === `file://${process.argv[1]}`) {
  main().then((c) => process.exit(Number(c) || 0)).catch((e) => {
    console.error(`✗ ${e.message}`);
    if (process.env.DEBUG) console.error(e.stack);
    process.exit(1);
  });
}